They are as incapable of handling one third of a dollar as binary positional notation is incapable of handling one fifth (0.2).

It’s not really a float problem. It’s a positional notation one. Some perfectly rational numbers refuse to squeeze into that mold.